usgs next generation topographic map - graphics research

USGS Next Generation Topographic Map - Graphics Research

West Virginia GIS Technical Center

Final Report

December 2009

Prepared by:

Kevin R Kuhn

Evan Fedorko

West Virginia GIS Technical Center Morgantown, WV 26506 304.293.9462

1

Contents i. Purpose ..........................................................................................................................................4

ii. Sections ................................................................................................................................................. 4

Section 1: Urban Tint Generation ......................................................................................................5

1.0.0 Urban Tint Concepts .................................................................................................................... 6

1.0.1 Data Sources ................................................................................................................................ 6

1.1.0 Urban Tint from Roads ..................................................................................................................... 8

1.1.1 Data .............................................................................................................................................. 8

1.1.2 Density ......................................................................................................................................... 9

1.1.3 Focal Statistics ............................................................................................................................ 11

1.1.4 Reclassify .................................................................................................................................... 13

1.1.5 Cartographic Processes .............................................................................................................. 13

1.1.6 Research Summary .................................................................................................................... 14

1.2.0 Urban Tint From Structures ........................................................................................................... 15

1.2.1 Data ............................................................................................................................................ 15

1.2.2 Density ....................................................................................................................................... 15

1.2.3 Focal Statistics ............................................................................................................................ 18

1.2.4 Reclassify .................................................................................................................................... 18

1.2.5 Cartographic Processes .............................................................................................................. 18


1.3.0 Urban Tint from Roads and Structures .......................................................................................... 20

1.3.1 Process ....................................................................................................................................... 20

1.3.2 Alternate Tint from Intersect and Merge .................................................................................. 20


1.4.0 Cartographic Process Steps ............................................................................................................ 23

2

1.4.1 Raster to Polygon ....................................................................................................................... 23

1.4.2 Remove Polygons ....................................................................................................................... 23

1.4.3 Aggregation ................................................................................................................................ 23

1.4.4 Smooth ....................................................................................................................................... 25

1.4.5 Clean / Erase .............................................................................................................................. 25

1.4.6 Technical Review ........................................................................................................................ 25

1.5.0 Final Conclusions ............................................................................................................................ 26

1.5.1 Challenges of Automation .......................................................................................................... 26

1.5.2 Research Findings ...................................................................................................................... 29

1.5.3 Further Research ........................................................................................................................ 30

1.6.0 Demonstration Topographic Maps ................................................................................................ 31

Section 2: Forest Cover Tint Generation ......................................................................................... 32

2.0.0 Forest Tint from NLCD .................................................................................................................... 33

2.1.0 Data Source .................................................................................................................................... 33

2.2.0 Process ........................................................................................................................................... 34

2.3.0 NLCD Process ................................................................................................................................. 34

2.3.1 Reclassify Raster ......................................................................................................................... 34

2.3.2 Raster to Polygon ....................................................................................................................... 34

2.3.3 Remove Polygons ....................................................................................................................... 35

2.3.4 Smooth Polygon ......................................................................................................................... 35

2.3.5 Remove Islands (Optional) ......................................................................................................... 35

2.3.6 Clean .......................................................................................................................................... 36

2.4.0 Generation Challenges ................................................................................................................... 36

2.4.1 Data Currency ............................................................................................................................ 36

2.4.2 Manual Interpretation ............................................................................................................... 36

2.4.3 Delineation ................................................................................................................................. 36

3

2.4.4 Data Resolution .......................................................................................................................... 37

2.5.0 Research Summary ........................................................................................................................ 38

Section 3: Structure Symbology ...................................................................................................... 39

3.0.0 Structure Symbology ...................................................................................................................... 40

3.1.0 Simple Structure Symbology .......................................................................................................... 40

3.1.1 Structure Point Rotation ............................................................................................................ 40

3.2.0 Complex Structure Symbology ....................................................................................................... 41

3.2.1 Structure Simplification ............................................................................................................. 41

Section 4: Appendix - Digital Processes Technical Specifications ................................................... 42

Technical Appendix ..................................................................................................................................... 43

4.1.0 Generation Process: Urban Tint from Road Data ...................................................................... 43

4.1.1 Generation Process: Urban Tint from Structure Data ............................................................... 44

4.1.2 Generation Process: Urban Tint from Road and Structure Data ............................................... 45

4.1.3 Cartographic Process ................................................................................................................. 46

4.2.0 Simple Structure Symbology Process ......................................................................................... 47

4.2.1 Simplification Process ................................................................................................................ 47

4

i. Purpose

The West Virginia GIS Technical Center (WVGISTC) at West Virginia University, Morgantown,

West Virginia, has completed its investigation for "Development of Automated Graphics Features for

Next Generation USGS Topographic Series in Support of the National Map." This report documents the

findings investigated for the scope of the project.

Since 1879, 1:24,000 scale topographic maps published by the United States Geologic Survey

(USGS) have been the gold standard of reference maps for the United States and its territories. Recent

history has seen the rapid emergence of new mapping technologies, and in response to those, the USGS

has developed “The National Map,” an internet based framework application for the sharing and

distribution of up-to-date geospatial data. The days of mass printing of 1:24,000 topographic maps are

gone. However, the topographic map is far from irrelevant and remains a central focus of the USGS. This

focus has shifted to digital maps but the type of information contained on traditional topographic maps

remains relevant. This transition to digital mapping not only applies to the information itself, but how

that information is cartographically represented. Two areas of USGS cartographic tradition were the

focus of this research project; the red urban tint that is used to reflect built-up areas and the green tint

used as a forest cover. As these maps were originally generated manually by skilled cartographers, a

new methodology to create these in mass, using digital data and automated processes is needed.

Identifying suitable data and creating these processes were the main goals of this research.

The report focuses on use of available large-scale data from state sources for incorporation into

the graphics production and use of automated feature extraction and statistical generalization

techniques to develop the urban area (red tint) and forest cover (green tints) polygons represented on

traditional 1:24,000 USGS quadrangles and for the next generation of USGS topographic map series. For

urban designations, the project utilized 1:4,800 scale building centroids, building footprint polygons and

road centerlines. For forest cover, the project investigated various combinations of forest land cover

types associated with the National Land Cover Database (NLCD 2001) and techniques to improve its

usability in a cartographic quality application.

This report and the research performed were based on several USGS technical documents. They

are referred to throughout this report as NMP, the abbreviation of the National Mapping Program

Technical Instructions (NMP) and Section Part. Specifically, NMP Part 6 covers the Built-up Standards

used for the Urban Tint generation and Part 9 details covers the Vegetative Surface Cover used for

Forest Cover Tint generation.

ii. Sections This report is divided into three major sections. Each section describes separate areas of research

performed for this investigation. Section 1 is titled Urban Tint Generation, which discusses the methods

and data used to generate the USGS Built-Up Areas (red tint). There are two subsections that discusses

various approaches researched utilizing road data and utilizing structure data. Section 2 of this report

discusses research performed to identify the data and process to generate the USGS Forest Cover or

Woodlands (green tint). Section 3 discusses additional cartographic techniques explored for structure

data in a digital map environment.

5

Section 1 Urban Tint Generation

6

1.0.0 Urban Tint Concepts

The WVGISTC explored several methodologies for creating the USGS urban tint areas using digital

processes. The goal of this research was to replicate the original criteria that USGS has used to define

and collect urban tint areas. The key elements of creating an urban tint can be rendered down from its

cartographic origins (NMP Part 6; Box 1), to identify the correlating GIS elements in its creation using

digital processes. Those elements are structures (buildings), their associated use areas(parking lots, etc)

and road networks. Using these elements, we are able to create conditions that equate to a built-up

area, through representations of structures and road networks.

This research focuses on two distinct aspects of generating an urban tint layer; the first using

road centerlines and the second utilizing structure points. The concepts and procedures in this report

provide for a repeatable methodology useable by other entities. We discuss the positive and negative

outcomes of these processes.

This research used the built-up density criteria and area size criteria to identify candidate areas.

A candidate area refers to areas that meet at least one criteria of the capture conditions. For example,

the density rasters will identify the first criteria of a 30% built-up density and additional processing will

narrow those areas by identifying minimum area size. Further research is needed in order for dimension

criteria to be used; this topic is discussed in Section 1.5.5.c

USGS NMP nomenclature uses the terms Urban Tint and Built-Up Areas interchangeably. This

report will distinguish between the two; Built-Up Area refers to a physical condition as described in Box

1. Urban Tint is defined as the desired GIS data layer, or graphic features, which represent Built-Up

Areas. Box 1 contains the criteria used to define an Urban Tint and the conditions which it is collected as

a layer.

The processes described in this report may refer to automated graphics. Automated graphics is

the term used to describe the digital processes and resulting Urban Tint layer researched in this report.

During this research these digital processes were performed manually. However, future research could

develop programming code or scripting to automate these digital processes.

1.0.1 Data Sources

West Virginia is blessed with several current, highly detailed datasets that will allow for a largely

automated derivation of these graphic features. These datasets were created as part of the West

Virginia State Addressing and Mapping program's work and include photogrammetrically collected

planimetric datasets, including the road and building data used in this research (Figure 1). This data was

BUILT-UP AREA - An area of intensive use, with much of the land covered by structures.

The limits of BUILT-UP AREA are determined by the relative concentration of buildings and associated

intensive use areas, such as parking lots, and the existence of a systematic street pattern. Density of

building and associated use areas will vary from densely concentrated areas downtown, to moderately

concentrated residential areas where most of the property is developed.

BUILT-UP AREA is > 30% built-up, and is > 40 acres in size, and is > 660 feet along the shortest axis

Box 1 - Part 6, NMP Technical Instructions

7

collected from 2-foot natural color, leaf off photography flown in 2003. Other states may not have this

type of data available, but it is expected that a point structure dataset and detailed road dataset will

become available as states further develop their GIS spatial data infrastructure. Furthermore, as new

programs such as Imagery for the Nation, help produce current, detailed views of the landscape, the

methodologies outlined in this investigation will allow urban tint areas to be generated and updated

with new information.

Figure 1 - Planimetric Data

8

1.1.0 Urban Tint from Roads The first process tested in this research utilized road data to create an urban tint layer. WVGISTC

researched the process and tools needed to transform road centerline data into a cartographic polygon

(Urban Tint) which could then represent built-up areas. This process established a correlation between

road segments and built-up area. Because not all roads may represent a built-up area, the road

centerline features were assigned a weighted value based on feature type. This linear feature data was

transformed into a raster dataset based on line density and feature weight. The resulting raster was

then processed to refine and enhance candidate areas based on NMP criteria. This raster data was then

converted to polygon data for further refinement and cartographic enhancement. The process

described on the following pages was designed to closely mimic the conditions that USGS uses to collect

a Built-up Area and produce an Urban Tint dataset.

1.1.1 Data

In an attempt to provide a repeatable process, we chose to use a nationally available dataset; a subset

of the US Census TIGER/Line. The 2003 SAMB 1:4,800-scale road centerline geometry was incorporated

into the Census TIGER/Line data as part of a major revision of the TIGER/Line data in 2007-08. Census

TIGER/Line data is categorized by multiple use codes (MAF/TIGER Feature class codes; MTFCC). The

TIGER/Line data contains numerous features (such as political boundaries) not of use in this study, so

the MTFCC codes were used to subset transportation features used in this research. These codes were

also used to assign feature weights to the roads (Figure 2). In our study, we used the following feature

codes and assigned weights (See Section 1.1.2.b).

MTFCC codes used and weight assigned: S1100 (Primary Roads) = 0 S1730 (Alley) = 1 S1740 (Private Road) = 1 S1200 (Secondary Roads) = 2 S1400 (Local Roads) = 3 Though we used Census TIGER/Line data for

this research, these concepts are applicable to

virtually any road dataset. Testing determined that

the type of road plays a role in accurate urban tint

generation and care needs to be taken in creating a

valid set of input features. Specifically, many

transportation datasets contain features such as on-

ramps, miscellaneous roads, trails and railroads.

These features were excluded from the process as

they may not reflect the desired results. Trails do not usually represent built-up areas. Railroads may

physically represent developed or built-up areas, but are linear in nature, and therefore not useful in

defining specific areas. Rail yards could be applicable, but are specified in the NMP Part6 as areas that

are not collected for built-up areas.

Figure 2 - Census Classification Weight

9

1.1.2 Density

We explored road density to assess and identify

areas that comply with the NMP Part 6, 30% Built-

Up criteria to produce an Urban Tint. The line

density tool was used as a direct method of creating

a raster surface from the road data. This tool results

in a raster dataset with values of density measure:

linear unit/per area. Figure 3 shows the source road

data on top of its density raster. Low density areas

are shown in reds, medium density shown in yellows

and high density areas are shown in greens.

Road centerline data is only one dimensional

and therefore needs to be assigned an equivalent to

an area value. Physical roads in the real world do

have an area, measured as the length and width.

(Theoretically, detailed centerline road datasets

could have this width data; but for this research we standardized road width as part of the weight value

(discussed in Section 1.1.2.b). The physical surface area of the road would be considered built-up and

would count directly towards the 30% built-up criteria of the area. The built-up criteria also incorporates

parking lots and structures as well as roads. Therefore, an additional area value needs to be assumed

and assigned to the road segments. Conceptually, this value is not only considering the road length and

width, but also the use area surrounding the road on either side.

Testing has shown that the density of the urban core is often close or equal to the density of

abutting or adjacent corridors and therefore they are often incorporated into the candidate areas.

Conceptually, this makes sense as inner-city roads are not significantly closer together or farther apart

than suburban neighborhood roads. This will vary with different areas of development and is also

impacted by numerous other variables such as terrain, city code or political boundaries.

There are several factors that influence the outcome of the line density raster, including cell size

and feature weight. We discuss the impact of these two

variables in the next section.

1.1.2.a Cell size

Two cell sizes were tested: 10 meter and 30 meter. Figure 4

shows the 10 meter cell in grey, the 30 meter cell in tan and

the Census TIGER Roads on top. The 10 meter cell size

produced a finer resolution raster, as expected, which allowed

a finer edge to the urban tint. A potential drawback of using a

10 meter cell size would be longer processing time and heavier

hardware requirements for performing the process over large

areas. The process will run quicker using the 30 meter cell size,

but, the resulting raster has less information (shape and

Figure 3 - Density Raster and Roads

Figure 4 - Cell Size Comparison

10

boundary detail). This slightly effected the cartographic boundary of the urban tint. Both cell sizes are

suitable for creating a density raster, but a preference for 10 meter cell size is given, should computer

processing power allow. This research used a 10 meter cell size; subsequent variables and thresholds

may need to be adjusted if a 30 meter cell size is used.

1.1.2.b Road Weight by Class

Using Census TIGER road classes, we were able to

assign feature weights prior to creating the density

raster. A feature weight works by allowing some

features to be counted more often than other

features. Assigning a weight value based on the

roads class allows for a better correlation between a

road segment and the surrounding built environment.

The concept behind the weight scheme is

that not all roads equally represent the built-up

environment. In general, highways traverse both city

and rural areas and do not necessarily represent a

built environment. Secondary roads can pass through

urban areas, but can also often be rural; therefore

they may or may not represent built environments.

Local roads are generally found grouped together and

often form the urban road network. Local roads are

therefore most likely to represent built-up areas.

During preliminary tests, density rasters

were created from road features without a feature

weight. This results with highway features being

over-represented as dense areas. Figures 5 & 6

display a density raster generated from road data.

Figure 5 clearly shows the inclusion of a highway

running from top to bottom on the left of the graphic.

Figure 6 shows the same area, but uses road features

with assigned weighting. Figure 6 shows similar areas

of density, but excludes the highway. The high

density areas that result from highways was due

because most highways are represented by dual line

geometry, which will have a higher density than

single line roads. High density which results from highways may often be a false representation. Typical

highways would not necessarily reflect a high density area, especially for rural states such as West

Virginia. Highways in general were not found to be beneficial in forming clusters of density. Even in

dense urban areas, local roads can better represent built-up areas than highways. In addition, the

linear nature of highways resulted in elongated, narrow, high density candidate areas. For these

reasons, highways were ultimately removed from inclusion when creating the density raster.

Figure 5 - Road Density Using Non-Weighted Roads

Figure 6 - Road Density Using Weighted Roads

11

For this research, we assigned weight values to features based on the Census TIGER/Line MTFCC

codes. The following classifications and weights were assigned: miscellaneous roads-1, secondary

roads- 2 and local roads-3 (See Figure 2). These weight values could be altered depending on the data,

criteria or area of interest. For example, alleys typically only exist in urban areas and could be more

heavily weighted than state routes.

1.1.3 Focal Statistics

The density raster is then processed through a

focal statistics function. The focal statistics

process refines and enhances the candidate

areas by maximizing the larger areas and

minimizing the smaller areas. The focal statistics

function used is a statistical mean; a statistic

that averages the density of a given area (the

neighborhood). The neighborhood is defined by

the number of cells of the raster that is used to

calculate the new value. Focal statistics was able

to highlight clusters and minimized the linear habit of the road density. This step is done in part to

accommodate the NMP Part 6 Specification to smooth and simplify the Urban Tint boundary (Box 2).

The size and shape of the neighborhood function effected the shape and composition of the

output candidate areas. In general, the larger the neighborhood size, the more smooth (or averaged)

the raster becomes. This was found to be desirable as it removed small undulations of the urban area

boundary and enhances key areas (such as larger areas of high density). This process was in keeping

with the NMP Part 6 (Box 2).This process must balance between a neighborhood size large enough to

minimize the smaller areas, yet keeping the neighborhood size small enough to limit the smoothing

effect.

Figures 7-11 demonstrate the effect

of altering the Focal Statistics neighborhood size and

shape. All figures shown were generated from the

same density raster and show the corresponding

source road data on top. The figures show the

reclassified focal rasters using the same classification

threshold, for sake of clarity. The classification

threshold is the cell value used to classify cells into a

candidate area class or non-candidate area class.

Section 1.1.4 will discuss this topic further. Figure 7

shows the original density raster (grey tint), which

has not been processed through focal statistics. The

figure shows that the density raster has several small

island areas apart from the larger areas and exhibits a

rough and jagged boundary. These traits are reduced

from using focal statistics process.

The outline of BUILT-UP AREA should be

relatively smooth. It should not be extended to

include a few buildings nor to exclude a few

buildings when the smoothness of the edge would

be effected.

When the perimeter streets are >30% developed,

the BUILT-UP AREA should be extended across

the street to include these buildings.


Figure 7 - Original Density Raster

12

Figure 8 shows the product of a rectangular neighborhood shape. The figure shows that a few

of the smaller islands have been removed; this shows the effect of lowering their cell values below the

reclassification threshold. Testing found that the annulus (donut) neighborhood shape tended to

smooth the boundary areas more than the rectangular shape. Figure 9 demonstrates the focal process

using a similar sized neighborhood, but in an annulus shape. This figure shows an even greater

smoothing effect and further reduction of small island areas. Neighborhood size also effected the shape

and area of the output candidate areas. The larger the neighborhood size, the greater the smoothing

effect it has on the output raster. Figure 10 shows a neighborhood radius of 30 cells (300m) than the

Figure 9 neighborhood radius of 15 cells (150 meters). Figure 11 shows a smaller neighborhood radius

using only a 3 cell radius. This figure demonstrates that few of the island areas have been removed and

only slight smoothing has taken place. This research used the annulus neighborhood shape, and a 2 cell

inner radius and 15 cell outer radius (Figure 9).

Figure 10 - Focal Raster, Annulus Neighborhood Shape, 2 cell Inner radius, 15 cell Outer radius



Figure 8 - Focal Raster, Rectangle Neighborhood Shape, 15 cell x 15 cell search size

13

1.1.4 Reclassify

Once the focal statistics raster has been calculated, the resulting raster (Figure 12) is then reclassified

into two data classes. This reclassification step acts as a threshold to determine which cells are

classified as candidate areas and which cells are not classified as candidate areas. The reclassification

plays an important role in determining the size of the overall candidate areas. Altering this threshold

can expand or shrink the candidate areas. The product of this reclassification step is another raster

which is then converted to a polygon dataset.

Our classification threshold value was 1.9. This threshold was chosen because it represented

the best balance of inclusion versus exclusion. The resulting classification shown in Figure 13, identifies

areas of high density with values of 1.9, shown as yellow. The 1 values, shown in green, and Null values

(white) are discarded. After several tests, it was found to be beneficial to be more inclusive during this

step. Other processes used later can be used to exclude areas.

1.1.5 Cartographic Processes

After the classification step, both the road generated and structure generated tint rasters undergo the

same process. Section 1.4.0 describes in further detail the processes used to clean and enhance the

data to complete the urban tint layer.

Figure 13 - Reclassified Road Density Figure 12 - Road Density

14

1.1.6 Research Summary

This research demonstrates that existing data and

methodologies can be used as a viable means to

bridge the past paper topographic maps with the

next generation of digital mapping products. The

road generated urban tints have stayed true to the

original USGS urban tints over the majority of the

study area, though there are differences. Most areas

of difference can be attributed to areas of genuine

new growth, as reflected in the source data, or

original tint areas that were interpreted by a

technician for inclusion.

Figure 14 shows the generated tint outline in

orange on top of a mosaic of USGS topographic

maps. This graphic illustrates that the majority of the

generated polygons coincide with existing USGS

urban tint areas. There are additional areas shown in Figure 14; these were included because they met

the same criteria that that is used to define the larger areas. These additional areas could be removed

by raising the threshold for minimum polygon size or by altering the aggregate distance.

Figures 15 & 16 show a comparison between the generated urban tint areas (Figure 15) and the

same area from a 1994 USGS 7.5-minute Topographic Map (Figure 16). The generated urban tint

reflects very similar boundaries to that of the original Urban Tint. The road generated tint includes areas

such as "Brewer Hill" (upper left of maps) and yet exclude areas such as "Suncrest Park" (center of

maps). Problems identified with the generation process includes area composition, shape and

boundary. These problems are discussed in detail in Section 1.5.1. Additional visualization of the urban

tint generated from road data can be seen on VERSION A of the Demonstration Maps.

Figure 16 - USGS Topographic map Figure 15 - Urban Tint Generated from Roads

Figure 14 - Generated Tint outline and USGS Topographic map mosaic

15

1.2.0 Urban Tint From Structures The second approach to generate Urban Tint utilizes

structure data consisting of building center points and

polygons (Figure 17). The concept behind using

structure data to create an urban tint is similar to the

use of road data to generate urban tint. A collection

of points (structures) can be correlated to the built-

up areas and be used to generate the urban tint layer.

The process of generating urban tint from structure

points is similar to the process used to generate

urban tint from road data.

1.2.1 Data

This section of the research primarily used point

structure data to identify candidate areas. In addition

to the point data, polygon structures were used to

enhance the structure point data. The structure data

was collected in 2003 as part of the WV Statewide

Addressing and Mapping program. All structures were collected with a center point. In addition to the

points, building footprints (polygons) were also collected for structures greater than 7500sq ft. Figure

17 shows the source data on the NAIP 2007 aerial photography.

It is recognized that not all states currently have structure point data and fewer areas will have

building polygon data. However, structure datasets are recognized as a key data layer and it is

anticipated that structure data will become widely

available as more states develop their GIS data

infrastructure. For this research, both point data and

polygon data was used to create the urban tint,

although this process can be used effectively with or

without polygon data.

1.2.2 Density

The point density process analyzed the relationship,

between the point data (structures) and how many

occur in a given area(statistical neighborhood), and

results in a density raster, as shown in Figure 18. We

used the USGS NMP Part 6 (Box 1) and directly

translated that criteria to GIS, utilizing mathematics

to model built-up areas.

Figure 17 - SAM Structure Point and Polygons

Figure 18 - Density Raster from Structure Points

16

In order to use points to create a

density raster, we made assumptions

about how much area the point

represents. In a residential area that is

considered built-up, a house sits on a lot.

This lot is considered to meet the conditions of the NMP Part 6 (See Box 4). As this definition explains,

the concentration of structures can vary within defined built-up areas. Residential areas would have a

single structure surrounded by an open space or yard, which would be represented by points spaced at

a distance. Theoretically, residential areas would be less dense than urban core areas. In an urban core

area, structures may be adjacent to each other, represented as points spaced closely together.

However, Figures 19-24 demonstrate that this generalization may not always be the case. These

figures show NAIP aerial photography on the top row, and their corresponding density grids on the

bottom row. All images show the source data structure points for reference. Figure 20 outlines two

areas for a comparison between an urban area (upper left outline) and a residential area (lower right

outline). Figure 23 shows the un-weighted point density raster of the same area. The color saturation

represents density; red colors are low density and green colors are high density. Figure 19 shows the

urban core area and the resulting density map (Figure 22). In contrast, Figure 21 shows a residential

area that is more dense (Figure 24) than the outlined urban area. The point density for these two areas

do not reflect the intuitive and expected result; that an urban area will have a higher density than a

residential area. There are multiple factors that determine why the urban point density is lower. As

Density of building and associated use areas will vary

from densely concentrated areas downtown, to

moderately concentrated residential areas where most

of the property is developed.


Figure 19 - Urban Points Figure 21 - Residential Points

Figure 22 - Urban Density Figure 23 - Comparison Site Density Figure 24 - Residential Density

Figure 20 - Comparison Sites

17

seen in the images, the building size of the urban structures are much larger than that of the residential

areas. In addition, the associated use areas of the urban structures, such as parking lots, are also larger

than that of residential use areas (such as yards). This information suggests that point data will need to

be weighted in order to better define built-up areas .

Even with the variability of density between urban and residential areas, the designation of an

urban tint area can be rendered down to a minimum density threshold. Urban and residential areas

that meet this minimum threshold would be candidate for inclusion in the urban tint, while areas that

fall below the minimum threshold would not be included. Using a minimum threshold based on an un-

weighted density grid will have varying results. Adding a point weight value was one solution used to

correct for the density disparity between urban and residential points, and the difference between

structure sizes.

1.2.2.a Point Weight by Structure Size

Similar to the road weighting, we looked at assigning weight

values to the building points before they are used to create the

density raster. The need for weighting the point data became

evident after initial tests showed that certain areas populated

with large buildings were being under-represented (Figure 25).

For example, two points that represent a residential house of

2,000 sq ft and office building of 100,000 sq ft have the same

effect on the density raster. The need behind assigning weight

values to particular point features reflected the disparity

between the density of large structures in the urban core and

generally smaller residential structures. Urban core areas

often have very large buildings, yet they can be adjacent (high

density) or spaced further apart (low density). Strip malls or

similar office complexes often contain large structures, parking

lots and associated use areas, but because their representative

points are spaced further apart, they result in a lower density.

This poses a problem when defining the area that would

normally be considered built-up.

Assigning a weight value to a point allows it to

represent a higher point density (Figure 26) than if it were not

weighted (Figure 25). We used a weight scheme that

correlated the size of the building footprint with the average

value of a single point. An average point represents a

structure (house and garage) and its associated use area (yard

and driveway), which in total represent 7,500 sq ft of built-up

area. This number was slightly less than what the US Census reports as the average lot size in a

Metropolitan Statistical Area (8,526 sq ft) sold in 2008. This also works out to be between 1/8 and 1/4

of an acre. Points which represents 7,500sq ft assumed an un-weighted base value of 1. For structures

larger than 7,500 sq ft, a weight value was assigned that equaled 2.5 times their measured square feet

Figure 25 - Density Un-weighted Points

Figure 26 - Density Weighted Points

18

value divided by 7,500. This formula directly relates the building footprint size value to the average

point value. The 2.5 multiplication factor is intended to account for the use area (parking lots) that the

building may have, which is not directly reported in the building's footprint size.

1.2.2.b Point Weight Alternative Value

Ultimately, the assigned weight value to a point is a means to better relate the point to the

actual structure and use area it represents. In this research, we established an average point value and

then assigned higher values to selected points based on structure size. This provides a way to

differentiate between points that would otherwise be valued the same interpolated through the density

raster. However, this type of weighting can be achieved without knowing the precise size of the

building. The principal is the same, regardless of which criteria is used. Zoning, deed or tax information

or simply a building type could all allow a similar type of point weight.

For example, zoning could be used as a criteria to classify building points. At a minimum, it could

be used to distinguish between residential, commercial and industrial buildings. For example,

residential points may represent 7,500 sq ft, commercial points may represent 25,000sq ft and industrial

points 50,000sq ft. This basic classification may be enough to differentiate how the point relates to the

actual structure and associated use area. In relation to the density map, the residential areas would be

held to a higher threshold than commercial or industrial areas. These types of adjustments can be made

according to region of the country or data availability.

In addition to zoning, some structure datasets may be attributed with a structure type, such as

"residential", "warehouse" or "retail". Almost any type of classification could be applied and used to

establish a weight scheme according to the availability of the dataset. More detailed classification of

weight values assigned will produce a point density raster that more accurately reflects the real world.

1.2.3 Focal Statistics

Unlike the road data, the point data did not need to be processed through focal statistics. This is in part

attributable to the geometry type. It is easier to extract clusters from the point data density raster than

from the road density raster.

1.2.4 Reclassify

The resulting density raster is then reclassified into two data classes. This reclassification step acts as a

threshold to determine which areas of density are included as candidate areas and which are not. The

reclassification plays an important role in determining the size of the overall candidate areas. Altering

this threshold can expand or shrink the candidate areas. The product of this reclassification step is a

raster which is then converted to a polygon dataset. For this research, our classification threshold value

was 2. This threshold was chosen because it represented the best balance of inclusion versus exclusion.

After several tests, it was found to be beneficial to be more inclusive during this step. Other processes

used later can be used to exclude areas.

1.2.5 Cartographic Processes

After the classification step, both the road generated and structure generated tint rasters undergo the

same process. Section 1.4.0 describes in further detail the processes used to clean and enhance the

data to complete the urban tint layer.

19


This research demonstrates that it is a feasible to generate urban tint areas from structure point data.

There are drawbacks to using un-weighted point data. However, when weighted building point data or

polygon data was used for enhancement, the process had very positive results. Most generated areas

were congruent with current topographic urban tint areas. Most areas of discrepancy were areas of

change, growth or additions. Other areas of differences noted could be attributed to manual

interpretation of the original tint layer. Figure 27 shows the generated urban tint and Figure 28 shows

the same area of the USGS topographic map.

The urban tint generated from point data has similar problems to the urban tint generated from

road data, though to a lesser extent. Of the problems identified with composition, shape and boundary,

boundary problems appear to be the most difficult to overcome. Using structure point data can result in

a more accurate representation of built-up areas than roads. The point density can be refined at a

tighter level than the road data. This is most likely due to the greater variability of point data compared

to the road data. In particular, point data density is well suited to identify clusters, while the road data

(linear data) density has a tendency of identifying corridors. Using point data can help to eliminate

protruding corridors and more closely represent the shape of the built environment. Point density does

have unique problems. For example, point density may identify a trailer court as having a higher density

than an urban core.

Unfortunately, even weighting has limitations. The point weight value is determined from the

size of the building footprint, not the actual size of the building. Therefore the weight value does not

distinguish between a one story building (such as a big box store) or that of a multi-story building (as

could be found in the urban core). Conceptually, this weighting could use actual square feet if that

information is available. Point weighting also tends to over concentrate the actual location of the

structure density. This is because a weighted point is only counted in a single cell which it overlaps, it

Figure 28 - USGS Topographic Map (Morgantown North) Figure 27 - Urban Tint generated from structures

20

does not distribute density to multiple cells. In reality, the actual building may cover many cells, in

addition to the structures use areas. This exhibits itself in the density raster as a rise in density value in

the shape of the neighborhood size, centered on the weighted point.

1.3.0 Urban Tint from Roads and Structures As this report has shown, there are limitations to how the urban tint is generated by both road data and

point data. However, one final aspect of creating an urban tint can utilize both road and structure data.

This third approach was explored to test the output of a process that uses the structure point and

polygon data in conjunction with the road data. There are several approaches that use both road and

structure data which produce varying results. For this research, WVGISTC altered the generation

process to utilize both road and structure data together. For sake of comparison, WVGISTC also explored

two basic GIS processes to combine the individually generated urban tint layers, discussed later in

Section 1.3.2.

1.3.1 Process

The overall process for utilizing both road and structure data is similar to their independent processes.

The difference is that both datasets were combined during the raster steps using map algebra before a

final reclassification threshold was chosen. Data integration earlier in the process allowed both data to

better influence the shape of the candidate tint areas. In addition, each dataset can be refined to

reflect its strengths during the density weight and classification process steps.

The process begins by individually creating a density raster from each dataset. The two rasters

are then individually reclassified. The reclassification gives a common value to both rasters, which are in

different density units. The reclassification assigns a higher value to areas of high density and a lower

value to areas of low density. The road density is reclassified using Natural Breaks, with four classes

while the structure data is classified using Natural Breaks with six classes. This un-balanced classification

was chosen to give preference to the structure data, which helps to limit area protrusions introduced by

the road data. The two reclassified rasters are then combined using raster math; a function which adds

the values of each cells in both rasters and produces a raster with the sum values. The resulting raster is

then reclassified again into two classes. For this research, we kept values of 5 and greater, values less

than 5 were discarded.

1.3.2 Alternate Tint from Intersect and Merge

For comparison, WVGISTC explored two standard GIS functions that would utilize both road and

structure tint layers to create a combined layer. After the urban tints are generated individually, they

can either be merged or intersected. The following figures demonstrate the individual product of all

three processes for comparison. These figures also display the original boundary of the individual layers.

The purple boundary represents tint area generated from road data, orange outlines the tint area from

structure data and green outlines the tint generated using both road and structure data.

When the two tint layers are merged, the net result is an increase in area. Figure 33 shows the

product of the merge function, which incorporates all areas from both tint layers. Almost as an opposite

process, the intersect function uses both tint layers to produce a tint layer containing only areas

21

common to both road and structure tint layers. Figure 34 shows the net result of the intersect function

as a subtractive process and produces a much smaller area than either individual layer.

Figure 35 shows the same area, but displays

the tint generated using combined road and

structure data. The figures show the outlines so that

you can see the areas of addition or subtraction

clearly. If you compare the green outline, it tends to

balance between the purple (road) and

orange(structure) areas. It is clear to see that by

using both data sources during the generation

process, the end product is not purely additive or

subtractive. For example; the center of the figures

show a cemetery, which has a high road density but

does not have structures. By utilizing both datasets

for the tint generation, the cemetery area was

excluded, as shown in Figure 35.


Using road and structure data for urban tint generation results in a tint that has a more refined area

composition and boundary. While areas that are mutually exclusive in both datasets can still be included

in the final data, this approach produces a compromise between the road and structure generated tint

layers. The generated urban tint appears to strike a balance between inclusion and exclusion of areas of

similar density within their respective datasets. This process allows for both data types to contribute to

the final shape and area of the urban tint raster. Initial tests show that by utilizing more information

results with a better defined urban tint area. This research has shown that it is feasible to closely match

the NMP criteria of using structures and road networks to establish areas that are built-up. These

Figure 35 - Generated Tint from Road and Structure Data

Figure 33 - Merged Road and Structure Tints

Figure 34 - Intersected Road and Structure Tints

22

processes can accomplish the task of producing an urban tint based on both structure and road density

data. Though the practicality of this process may not be initially viable due to lack of data, as additional

data becomes available, this research may be utilized in producing the next-generation of map layers.

Figures 29-32 display the same area of the Lake Lynn Quadrangle at 1:24,000 scale. Figure 29

displays tint generated from road data, which results in the largest tint area. This figure also shows how

the tint follows the road corridors on the upper left and top of the map, where the tint continues off the

map. Figure 30 shows tint generated from structure data. This figure shows a smaller area than the

USGS tint (Figure 32), but shares very similar boundaries overall. Figure 31 shows the tint generated

from both road and structure data. This figure shows a similar boundary to the USGS tint area, however

has included less area towards the bottom than the structure tint. Figure 32 shows the original USGS

topographic map. Though all three generated tints do not match the original USGS tint, they do all

convey a close approximation to it. For further visualization of the urban tint generated from both road

and structure data see VERSION C of the demonstration maps.

Figure 31 - Generated Tint from Road and Structure Data

Figure 30 - Generated Tint from Structure Data Figure 29 - Generated Tint from Road Data

Figure 32 - USGS Topographic Map (Lake Lynn)

23

1.4.0 Cartographic Process Steps Although the method of generating candidate urban tint areas may differ, there are similar processes for

generating the cartographic output. The cartographic process begins after the density rasters have been

reclassified into a two class, integer raster. This reclassified raster is then converted to polygons for

further refinement. Any areas of variance from standard procedures will be noted when appropriate.

1.4.1 Raster to Polygon

This step converts the reclassified raster data into a polygon layer. The conversion to a polygon layer is

necessary for a number of reasons. Many cartographic enhancement processes and end use of the tint

layer require a vector dataset, not a raster data format. For this research we converted the raster to the

Shapefile format, though a geodatabase feature class could be used. This step used the reclassified

values to create a polygon layer attributed on class value. During this step, we remove all polygons of

class value of 1 (See Figure 13). One variable to note using the ArcGIS tool "Raster to Polygon" is the

option to simplify polygon features. This option is not used in this step to allow for a separate smooth

step later.

1.4.2 Remove Polygons

This step removes polygons using a threshold

based on the size of the polygon. This step is

based on the NMP Part 6 (Box 5) minimum area size. Figure 36 shows areas over 40 acres in purple and

areas under 40 acres in red.

This step was performed in two different orders during the process; before and after

aggregation. Changing the order of this process step has consequences that effect the composition of

the candidate areas. See section 1.4.3 for further explanation.

1.4.3 Aggregation

This step performs an aggregation function on the

candidate urban area polygons. This step is based

on the Part 6 Technical Specification (See Box 6).

The aggregation step accomplishes several tasks;

first it combines adjacent polygons that are within

660 feet of each other.

This step also removes holes of certain sizes

from within polygons. For this research we set a

criteria to fill holes less than 3,000 acres. There did

not appear to be corresponding USGS

documentation regarding the need for a continuous

urban tint. Using existing quads as examples, this

step seemed appropriate.

Finally, this step removes polygons under a Figure 36 - Candidate areas < 40 acres in red.

Capture Conditions: If BUILT-UP AREA is >

30% built-up, and is > 40 acres in size, and is >

660 feet along the shortest axis, then capture.


24

certain threshold based on area size. Our threshold was set at a minimum of 40 acres to correspond to

the capture conditions. This part works in conjunction with the previous step, regardless if polygons

under 40 acres were already removed.

As noted in section 1.4.2,

removing undersized polygons prior to

the aggregation step is thought to

adhere better to the current USGS

Technical Specifications. The NMP Part 6 (Box 6) refers to extending larger built-up areas to incorporate

smaller built-up areas. This was interpreted to exclude areas that would fall below 40 acres

independent of the larger built-up area. By removing these areas prior to aggregation, they will not be

included in the urban tint area. When candidate areas < 40 acres are removed before aggregation, the

overall remaining polygons will generally have a smaller number of inclusions and therefore cover a

smaller overall area.

For this research, the removal of areas < 40 acres was performed before aggregation. Figure 37

shows areas < 40 acres outlined in red and the resulting tint when they are not removed prior to

aggregation. This shows the additional areas of inclusion that have been aggregated into larger polygon

areas. Figure 38 shows the same extent; areas < 40 acres are shown in red, and the resulting area when

they are removed before aggregation. Removing polygons prior to the aggregation step may initially

seem positive, but there are drawbacks to this. For example, there are areas that should be

incorporated to generalize the boundary, yet are excluded. There are also multiple groups of smaller

areas that may total greater than 40 acres when collected as a whole; these areas may traditionally be

collected. However, if they individually fall below 40 acres, they are removed and therefore excluded

from incorporation. This problem is mitigated somewhat by the focal statistics process (for the road

tint), but as noted, this process must balance between allowing islands and overt smoothing of the

boundary.

Figure 37 - Areas < 40 acres included in aggregation Figure 38 - Areas < 40 acres removed prior to aggregation

A large BUILT-UP AREA that is separated from a smaller

isolated BUILT-UP AREA, may be extended to include the

smaller area, if the smaller area is within 660 feet of the

larger BUILT-UP AREA.


25

1.4.4 Smooth

The step removes the pixel edges and smoothes the polygon boundary for a cartographic quality layer.

Figure 39 shows the polygon before processing; note the pixilated boundary edge. Figure 40 shows the

smoothed edge and Figure 41 shows the same area in context with other map layers. The variable of

smooth tolerance is a balance between an over-simplification of the polygon edge and inclusion of

detail, such as the end of a road. For this research, we used a smooth tolerance of 120 meters for both

road and structure tint layers.

1.4.5 Clean / Erase

The final process step cleans the urban tint polygons of certain areas such as rivers, lakes and parks

from the generated urban tint layer. This step is to comply with NMP Part 6 (Box 7) criteria for areas not

to collect as urban tint. This step is necessary for several reasons. The density raster used to create the

candidate areas ignores specific boundaries such as rivers and lakes and the resulting tint polygon may

overlap these areas. In addition, small areas such as a city park may be filled in during the aggregate

step. The clean process uses the erase tool to subtract undesirable areas of overlap from the final tint

layer. All of these layers may not be available as polygon data for use as an erase layer, however larger

areas such as rivers and lakes should be widely available. Additional data can be used at any time after

this step to perform this step and enhance the urban tints cartographic properties.

1.4.6 Technical Review

Ideally, the generated tint polygons would

undergo final inspection from a technician. This

inspection might compare new areas with older areas,

may clean or adjust certain polygon boundaries and check

against any blatant errors introduced through automation.

The result of this step would be the final Urban Tint

polygon for use in a digital mapping environment.

If the following features meet capture conditions for that feature, then do not collect BUILT-UP AREA: Lake/Pond Reservoir 2-D Stream/River 2-D Canal/Ditch Swamp/Marsh Railway Yard Park Athletic Field (other than tracks) Cemetery Golf Course Exhibition Ground Outdoor Theater Runway/Apron/Taxiway


Figure 41 - Generated Tint Figure 40 - Polygon Smooth Figure 39 - Polygon Pre-smooth

26

1.5.0 Final Conclusions

The WVGISTC investigated the use of structure data and road data as separate means to generate an

Urban Tint layer. This report has discussed the results of this research and demonstrated what can be

achieved using digital methods. This report details the process steps and decisions used to generate the

Urban Tint, as well as their potentially positive and negative results. It is intended that USGS will use

this research to help initiate implementation of some of these tools and processes in its digital mapping

products. The methodology using a weighted road dataset to generate urban tint is currently the most

usable approach, given the wide spread availability of road data. However, problems remain with a fully

automated process which may hinder the implementation of these digital processes without a review

step by a technician. Using a weighted structure dataset appears to produce a more accurate urban tint

that is ready for immediate use than an urban tint generated from roads. However, current availability

of structure data may pose a significant limitation for most areas. This may change as more structure

data becomes available, making the structure process and combined data process more feasible.

1.5.1 Challenges of Automation

This research has identified several key areas of difference between the digital process and the original

manual process. The issues that may pose problems to use are related to composition, shape, and

boundary. The tint layer composition refers to which areas are included or excluded during the creation

of the urban tint layer. Shape refers to the overall boundary (or shape) that the urban tint layer has

(such as circular, amorphous or elongated). The boundary issues refer to the large scale, high detail

edge that the tint area has and generally how it may perform cartographically. The following sections

describe the problems in greater detail, using Figures 42-45 to demonstrate example areas. The figures

shown include all three generated tint boundary outlines. The outline colors are: Purple Boundary Tint

from Roads, Orange Boundary Tint from Structures and Green Boundary Tint from Road and Structures.

1.5.1.a Composition

In Figure 42, a major area of difference can be seen

around "St. Francis High School" (middle right of

map). USGS may have originally included this area

due to the presence of multiple large structures in the

context of a large built-up area. The buildings are on

the campus of West Virginia University and are

connected by a single road. This type of

interpretation would be evident from aerial

photography, but not through the automated

process. This area was not included in the road

generated tint due to a low density of roads. The tint

from structures did include the area, but the

boundary is much more irregular than the USGS tint.

Although the generated tints can closely

approximate the built-up conditions, it is not able to

catch all possible conditions that might equate to the

Figure 42 - Generated Tint Outlines and USGS Topographic Map

27

original capture conditions. Tints generated from roads were susceptible to misclassified roads which

can incorrectly identify candidate areas. This was shown on Version A of the Lake Lynn Quad, which

incorporated two sections of incomplete highways that were classified as local roads. Other

composition differences can result from the traditional method of manual interpretation which

identified structures, roads and use areas in photography. Using the road and structure data alone to

determine density may excluded areas which should be included. For example, Figure 19 showed few

points in an urban environment, leading to a low density. However, the large parking garages and

parking lots were not represented by the structure data.

This problem could be overcome by using additional data, such as an enhanced weight scheme

or additional data to enhance or verify built-up areas. One possibility is to check identified candidate

areas against land cover data for verification. This is an area that needs further research.

1.5.1.b Shape

The generated tint area generally matches up

well with existing USGS urban tint areas.

However, the automated method has a

tendency to include protruding road corridors

that lead into and out of the urban area. This

is an undesirable trait as described in the NMP Part 6 (Box 8). These candidate areas often reflect the

urban build up surrounding larger urban areas, specifically road corridors adjacent to, but detached

from larger urban areas. This problem is more prominent in the Urban Tint generated from roads, but

the Urban Tint generated from structures also

showed this problem in a few areas as well.

Figure 43 shows the generated tint area

outlines on top of a USGS topographic map (faded

for clarity of boundary lines). This graphic

demonstrates how protruding areas are included in

larger areas. The figure shows that the tint area

boundaries all follow a road corridor out to a larger

built-up area. All areas within the outlines meet the

minimum threshold for creating an urban area and

are therefore included.

The generation process must strike a balance

between being inclusive of an urban areas edge,

without extending the boundary of the main area

which includes areas along a corridor. This is

difficult based solely on density. Therefore the shape

of the candidate areas must factor into the final urban tint area. Utilizing shape of candidate polygons

was explored by using a Minimum Bounding Rectangle script, but not fully researched in this study. This

is discussed further in Section 1.5.3.c.


Narrow strips where there is only a road and

buildings along both sides of the road should not be

included even if they would connect an adjoining

larger area that meets capture conditions.


28

Another possible solution could be to use algorithms that identify the core area of the candidate

areas, this information could then identify protruding areas outside the core area, which could be

excluded from the final Urban Tint polygons or marked for trimmed. This step could be included in the

automation process, however using any automated process to make these distinctions may leave an

undesirable or abrupt boundary. This is an area

for further research.

1.5.1.c Boundary

The third issue identified is the structure of the

edge or boundary of the generated urban tint.

Boundary delineations are ignored during the

generation of urban tint. The NMP Part 6 (Box 9)

states that the edge of urban tint areas should

follow and adhere to logical boundaries such as

roads, streams or other edge features. The

automated process is not able to consider such

cartographic determinations when the shape of the

polygon is constructed. Figure 44 shows how the

USGS urban tint has been constrained to a municipal

boundary to the right and bottom, and constrained

to a road towards the top. The generation process is

not able to reflect these types of cartographic

determinations. The three tint boundaries have

correctly identified this area as built-up, but are not

able to adhere to the strict boundary as cleanly as

the original USGS boundary.

The generated tint boundary may also

overlap areas that would seem illogical, such as a

river. The generated urban tint can overlap these

areas for several reasons; the source data may not

accurately reflect a built-up condition. An example,

of this may be a bridge (density over water) or a

cemetery (high road density, no structures).

Another reason may be the density raster or focal

statistics rasters were classified or smoothed, which

may cause high density areas to cover non-built-up

areas. A third possibility is that the aggregate

function may have incorporated an area that does

not meet the built up criteria, such as shown in

Figure 45. The areas on both sides of the river meet

the capture conditions, and they are within 660 feet

(the aggregation distance) and therefore are

Although Built-Up Area limits need not be fixed in

relation to linear features, where practical, the

limits of Built-Up Area should share the edge with

other linear features such as roads, railways,

stream/rivers, boundary lines, etc.




29

aggregated together to cross the river. The boundary of the generated urban tint areas show the areas

of greatest difference between the manual USGS process and this digital process. The edge differences

may also be the most challenging of these issues to overcome through automated means. One possible

solution could be to use topological tools that can alter the polygon boundary to snap to adjacent

features. This would require additional data sources, such as municipal boundaries or streams, but this

should not be a limiting factor. Ultimately the tint areas could meet the NMP guidelines by fully

automated methods, but further research in this areas is needed.

1.5.2 Research Findings

The USGS National Mapping Program Standards set forth in the Part 6 manual were intended to create

USGS Urban Tint areas as cartographic features for printed topographic maps. The original intent of

Urban Tint areas was to denote areas of dense, built-up areas and to signify generalization of building

structure symbology. Those same standards set to a digital process can result in areas that do not

reflect the spirit of the original urban tint area. However, this research has shown that the USGS

process and criteria used can be altered to create areas that more closely reflect the spirit of the

original urban tint, if not the exact areas. There are several areas which we have identified as candidate

areas of change, which may help transition traditional USGS methods towards automated digital

methods.

1.5.2.a Transitioning Standards

There are difficulties in producing a clean, cartographic boundary to the tint area. However, the

edge of the tint area may be less of an issue in a digital environment than on paper. Digital mapping

techniques such as scale dependency or data generalization may allow for the urban area to be

displayed with good cartographic principals. Print maps generated at 1:24,000 scale can still use the

newly generated urban tint, though the urban tint edge may not be as refined as the manual process.

These new urban tint polygons can be used to identify areas of high density, and therefore be used to

identify areas that may need to have generalized building symbology. The generated tint polygons can

be used to clip building point datasets for use as a scale dependant layer.

Preliminary research for an improved raster using both structure point and road data as an

enhanced integrated study to identify built-up areas resulted in improved urban tints. Though this

approach may be feasible in the future, it may be less practical now as structure data is not as widely

available as road data.

1.5.2.b Tint Generation Hybrid Approach

The process outlined in this report does present challenges; however, the majority of the automated

process results in a product which stays accurate to the goals of the project. One potential solution is a

compromise to both traditional cartographic techniques and use of automation. By using the processes

outlined in this research, USGS could generate new candidate areas as data sources become available.

Once these urban tint polygons are generated, there could be follow up from a technician using digital

editing tools to clean, or adjust the boundary where necessary. This could still save considerable time

from the manual process of identifying the urban tints. It is conceivable that the automated process

could be run on an annual basis or target designated areas for updating. This could be done as needed

or from a local National Map partner as data becomes available or workloads permit.

30

1.5.2.c Utilization

The urban tint that results from this process may have purpose in a digital mapping environment,

without manual processing. At moderate scales (ex. 1:24,000), the urban tint can be used to mask or

clip structure point data. This would allow the map reader a clearer look at the data and fulfills the

original spirit of the urban tint. Figures 46-48 demonstrates such an example. Figures 46-47 are digitally

generated maps displayed at 1:24,000 scale. Figure

46 shows an area of high structure density without

use of an urban tint mask, displaying all structures.

Figure 47 shows the same area with the generated

tint polygon (un-edited) used as a building mask. For

reference, Figure 48 shows a mosaic of the original

USGS Topographic maps. When compared to Figure

48, the generated urban tint on Figure 47 does not

match the original tint exactly. However, when

compared to Figure 46, it does perform the same

task as was intended from the original urban tint.

For instance, though these two maps do not display

any road annotation, it is quite clear which map

would be able to better accommodate road feature

annotation.

1.5.3 Further Research

This research has tested numerous variables that can affect the output of the generation process.

However, many additional tools and concepts remain untested. Additional processing steps may be able

to help reduce the problem issues or enhance the output of the generation process. Some of these

areas for further research are listed in this section.

Figure 48 - USGS Topographic Map Mosaic

Figure 46 - Generated Topographic Map

Figure 47 - Generated Urban Tint and Topographic Map

31

1.5.3.a Automation Script Development

This research performed the tasks outlined in this report through manual means. Automated scripts

could be developed to allow for a fully automated process, allowing large input datasets to be processed

by USGS as they become available. This automation could be used as an update tool when new data

becomes available. The scripting could be written using Python or Visual Basic and potentially operate as

a custom tool in ArcMap or Arc Catalog.

1.5.3.b Improved Boundary

One of the issues identified in this research is the boundary or edge of the generated urban tint. This

problem was not addressed through this research, but new research using custom tools or topology may

be able to improve or replicate the desired cartographic principals of the original urban tint.

1.5.3.c Improved Composition

As the automation process is developed, additional

tint criteria could be used for the selection of

candidate areas. One such criterion could be based

upon the shape of the area. Figure 49 shows one

such process which uses a script to identify the

minimum bounding rectangle of the area. The

minimum bounding rectangle information can then

be used to measure the overall width and length of

the area and give each polygon a length to width

ratio. Using this technique, it may be possible to

identify corridors or elongated areas that are

undesirable and remove them before they are

included in the final tint area. Initial exploration of

this technique appears to be promising, but the

script as written only considers the full width and

length of the bounding rectangle, not the actual

polygon. Further research is needed.

1.6.0 Demonstration Topographic Maps The demonstration topographic maps are meant to show the results achievable using digital processes

for Urban Tint Generation. There are four variations of three USGS quadrangles: Morgantown North

WV-PA, Morgantown South WV and Lake Lynn WV-PA. Versions A, B and C each display the Urban Tint

layer as generated from various data sources. Version A shows the Urban Tint generated from road

data. Version B shows the Urban Tint from Structure Data (both point and polygon). Version C shows

the Urban Tint resulting from using both road and structure data. Version D shows no urban tint; rather

it is meant to display all the road and structure data used to generate the Urban Tints. There was a

simplified process and single data source for the forest tint generation which resulted in one tint layer,

which is displayed on all three versions. See Section 2.0.0 for further information on the Forest Tint

Layer.

Figure 49 -Polygons shown by Length/Width ratio

32

Section 2 Forest Cover Tint Generation

33

2.0.0 Forest Tint from NLCD The green tint symbology that appears on the USGS 7.5 minute topographic maps is intended to

represent tree cover (Woodland). The NMP Part 9 (Box 1) describes the Woodland tint. This data has

traditionally been collected under photogrammetric standards that specify requirements of canopy

height, minimum widths and tree density which result in a green shaded area (polygon) that symbolizes

woodlands. USGS documentation refers to “Trees”,” Woods” and “Woodland” and for the purpose of

this paper all are considered the same. This section researched the method and data source to provide

for a suitable digital feature that

can be used with digital mapping

products.

The data most suitable for creating a forest cover tint using digital processes is the National

Land Cover Dataset (NLCD). Three classifications of the NLCD can be used to represent the forest cover

tint. This section of research discusses the process used to extract the useful data and create the tree

cover tint layer.

2.1.0 Data Source The NLCD is created and supplied by a federal program with well defined standards and

periodic, standardized updates. The NLCD is published by the Multi-Resolution Land Characteristics

Consortium (MRLC), which is a group of federal agencies formed to develop a national, uniform, land

cover dataset. The NLCD data is consistent across the nation making it well suited as a data source for

the generation of a forest cover tint. NLCD is published in a raster format which was created by an

unsupervised classification of Landsat data. The NLCD uses a modified Anderson Classification and has a

spatial resolution of 30 meters, originating from the Landsat 30 meter cell size.

In addition to forest cover, this data could be used to provide a more detailed tint layer, such as

showing distinct coniferous and deciduous forest areas. Though the NLCD may be coarse for some

applications, the NLCD could also be used to show orchards, cropland and wetlands where more

detailed data is not available.

TREES - An extensive area of land covered with woody perennial plants having a self-supporting main stem or trunk, a definite crown and usually reaching a mature height of more than 6 feet.


41. Deciduous Forest: Areas dominated by trees generally greater than 5 meters tall, and greater than 20 percent of total vegetation cover. More than 75 percent of the tree species shed foliage simultaneously in response to seasonal change. 42. Evergreen Forest: Areas dominated by trees generally greater than 5 meters tall, and greater than 20 percent of total vegetation cover. More than 75 percent of the tree species maintain their leaves all year. Canopy is never without green foliage. 43. Mixed Forest: Areas dominated by trees generally greater than 5 meters tall, and greater than 20 percent of total vegetation cover. Neither deciduous nor evergreen species are greater than 75 percent of total tree cover.

Box 2 - MRLC NLCD Classification Description

34

2.2.0 Process The WVGISTC used ESRI ArcGIS products for the process of creating the forest cover tint. In

summary, the NLCD data is reclassified in to two classes: NLCD Values 41,42,43 (which represent

Deciduous, Evergreen and mixed forest types) are assigned a class value of '1' and all other values are

assigned a class value of '0'. This reclassified raster is then converted into a polygon, where all '0' values

are removed. The remaining polygons then go through a smooth process. For the purposes of this

research, these processes were performed manually, however a script could be written to fully

automate the process according to the desired workflow. A detailed outline of the process is below.

2.3.0 NLCD Process

2.3.1 Reclassify Raster

This step separates out the usable data from the data not used

for creating the new tint. Figure 1 shows the Anderson

classified NLCD.

a. Tool: Spatial Analysis>Reclass>Reclassify

i. Input: National Land Cover Dataset

ii. Reclassify: Gridcodes 41,42,43 to ‘1’,

All other values to ‘0’

iii. Output: NLCD Reclassified

2.3.2 Raster to Polygon

This step converts the reclassified raster into a shapefile

polygon. Figure 2 shows the re-classified raster. Purple shows

forest areas, teal shows non-forest areas.

b. Tool: Conversion Tools> Raster to Polgon

i. Input: NLCD Reclassified

ii. No Simplify polygons

iii. Output: NLCD Polygon

Figure 1 - Classified NLCD Raster

Figure 2 -Reclassified NLCD Raster

35

2.3.3 Remove Polygons

This step removes the unused non-tint data polygons. Figure 3

shows the remaining forest cover polygons over aerial

photography for reference.

c. Input: NLCD Polygon

i. Process / Tool: Select

1. [gridcode] = ‘0’

ii. Process / Tool: Delete

1. Delete Selected

d. Output: NLCD Clean

2.3.4 Smooth Polygon

This process smoothes the polygon edges for cartographic

use. Figure 4 shows the un-smoothed polygon in purple and

smoothed polygon in green.

e. Tool: Data Management Tools > Generalization

> Smooth Polygon

i. Input: NLCD Clean

ii. Paek: 120 meters

iii. Preserve endpoints

iv. No Check

v. Output: NLCD Smooth

2.3.5 Remove Islands (Optional)

This step removes single pixel polygons. In this step, creating a

new field to add acreage information allows for selection and

removal based on area criteria. This step can be skipped to

include all areas. Figure 5 shows a single pixel polygon

highlighted in blue outline.

f. Input: NLCD Clean

i. Add “Area” field

ii. Calculate “Area”

iii. Select: Select < .22 acres

iv. Delete Selected

g. Output: NLCD Clean2

Figure 3 -Extracted Forest Class Polygons

Figure 4 -Original Polygon and Smoothed Tint

Figure 5 - Single pixel shown in blue outline

36

2.3.6 Clean

This step removes unwanted areas from the generated tint

areas. These areas would follow the NMP Part 9 Tree Cover as

areas that are not to be collected as Forest Tint. Figure 6

shows the green tint overlapping a hydrography layer. In this

example, these overlaps would be erased.

h. Tool: Analysis Tools>Overlay>Erase

i. Input: NLCD Clean

ii. Erase Input: Layer

i. Output: Final Tint

2.4.0 Generation Challenges The following are some potential issues that arrive from generating a forest cover tint layer from the

NLCD

2.4.1 Data Currency

The current NLCD was published in 2004 using Landsat imagery from 2001. The MRLC recognizes the

temporal nature of its land cover dataset and is currently working towards updating the NLCD to

include 2006 data. It could be assumed that if the NLCD tree cover would change, those changes would

likely be areas of subtraction, such as areas of urban growth. Because many areas of subtraction are

created and then processed independently of the tint layer, they could then be applied as they are

revised to the tint layer. Areas of new tree cover growth would be captured with updates of the NLCD,

and therefore would be incorporated into a revised tint as the NLCD is released.

2.4.2 Manual Interpretation

The NMP Standards state the following : “Depict forested areas that have been clear-cut, but are

expected to be reforested, as TREES.” Though this type of manual decision making has been removed

through automation, the problem may be less relevant with dynamic data as it was when creating a

paper based product. The update cycle of digital data allows for a more accurate and temporal layer.

2.4.3 Delineation

The NMP Standards state the following: “The limit of TREES is the extent of the treed area. The limit is the center of the outermost trees, not the edge of their crowns.” This interpretation is not feasible using this methodology. The NLCD 30-meter cell approximates tree density, but does not locate a specific crown or tree. Therefore, the generated edge will be only an approximate tree boundary. This may not be as accurate as a manual interpretation.

Figure 6 -Tint shown overlapping water

37

2.4.4 Data Resolution

The NLCD source data (Figure 7) is based on 30-meter Landsat Data. It could be assumed that a tint

generated from the NLCD directly would then appear pixilated or coarse, as shown in Figure 8.

However, with correct processing, this is not necessarily true. The process outlined in this report

incorporates a smoothing step to compensate for the pixel edges and produce a cartographic quality

layer. Figure 9 shows the resulting tint in context with other layers. Though the pixel edge has been

corrected by smoothing, the actual tint boundary may be more of an issue at larger scales. At these

detailed scales, the real edge of the tree cover may not match the symbolized tint. However, the forest

cover representation is still able to reflect the approximate tree cover location. Figures 7-10 depicts the

source data and resulting forest tint products at 1:4,800 scale. Figure 8 shows the source raster edges

with NAIP aerial photography shown for reference. Figure 10 shows the USGS Topographic Map for

comparison. These figures demonstrate how a smoothed tint area can act as a cartographic quality

forest tint.

Figure 7 - Classified NLCD Raster Figure 8 -Forest classes converted to polygon

Figure 9 - Generated Tint Figure 10 - USGS Topographic Map

38

2.5.0 Research Summary This research has shown that generating a new forest tint layer using the NLCD has worked exceptionally

well. The digital processes outlined in this report could be fully automated, allowing USGS to process

large areas. Figures 11-14 show a comparison between USGS Topographic maps and the generated tint

at two scales. Figures 11 & 12 displays the maps at 1:10,000 scale. Figure 11 shows the USGS

Topographic map. Figure 12 shows the corresponding area using a digitally generated map with the

NLCD generated forest tint. Figures 13 & 14 show the maps at 1:24,000 scale. Figure 13 shows the USGS

Topographic map, and Figure 14 shows the Generated map and tint. For additional visualization, please

see the Morgantown North, Morgantown South and Lake Lynn demonstration maps that use the

generated forest tint from this research. There are three versions of each quadrangle, however the

forest tint layer is identical on each map.

Figure 11 - USGS Topographic Map, 1:10k Scale Length/Width ratio

Figure 12 - Generated Tint and Map Layers, 1:10k Scale

Figure 13 - USGS Topographic Map, 1:24k Scale Figure 14 - Generated Tint and Map Layers, 1:24k Scale

ratio

39

Section 3 Structure Symbology

40

3.0.0 Structure Symbology This report describes the use of point features and polygon features to represent structure (buildings)

features that replicate USGS NMP Part 6 cartographic standards. As structure datasets become more

widely available, they are often being utilized on digital map products. USGS uses a distinctive set of

symbology to represent structures on 7.5 minute topographic maps. There are two types of generic

structure representation: a single, simple square that represents average structures and a more

complex shape that represents larger structures. This report is not meant to address specific structure

symbols such as churches or schools. There are two brief sections that addresses each of these

structure symbologies. Section 3.1.0 discusses the simple shape symbol and how to achieve a point

symbol rotation for alignment to road features. Section 3.2.0 discusses the simplification of detailed

polygon features for building representation in digital map products. These processes were performed

using the ESRI ArcGIS suite of software.

3.1.0 Simple Structure Symbology This process describes the application of using a point dataset as a representation of structures for use

in digital map products. The 7.5min series USGS topographic maps use a 2pt, black, simple block

symbol to represent generalized building locations outside urban tint areas (NMP Part 6-20 thru 6-31).

For cartographic aesthetics, the block symbols are rotated to align with its corresponding street. These

are procedures that replicate the structure orientation.

3.1.1 Structure Point Rotation

This process was done using ESRI ArcGIS – Arc Catalog Toolset using the Near tool. This process

requires a structure point dataset and a road dataset at approximately the same scale. The Near tool

will calculate the distance and direction to the closest road feature from each individual point feature.

This information is written into the point datasets attribute table. The distance will not be used and

may be deleted; the Near angle field will be used to specify the rotation angle for the features

symbology. When symbolizing the building point dataset, under the advanced properties, there is an

option to specify a field that contains the rotation angle. The Near angle field should be specified here,

and the option of Arithmetic should be chosen. Figure 1 shows the un-rotated 2pt block symbol. Figure

2 shows the same block symbols but rotated using the Near Angle added through this process.

Figure 2- Structures aligned to roads Figure 1- Structures not aligned to roads

41

3.2.0 Complex Structure Symbology Complex structure polygons can add detail to map products, but they can also make the map un-

necessarily complex and difficult to read. In addition, large structure datasets may hinder functionality

or performance of online digital map systems. Structure simplification reduces the polygons complexity

and can produce a smaller dataset which allows easier readability and greater performance. This

section describes the process and tool used to simplify building structures.

3.2.1 Structure Simplification

Within ArcGIS there is a specific tool called "Simplify Building". This tool processes the input polygons

and removes certain vertices, resulting in a simplified shape. The tool has two options: the

simplification tolerance and minimum area. Testing of the simplification tolerance showed a desired

range of 30 feet and 80 feet, depending on the desired level of simplification. Figure 3 shows the

original structure polygons. Figure 4 shows the simplified building with a tolerance of 80ft. Figure 5

shows the comparable USGS topographic map for reference.

Figure 3 -Original Structure Polygons Figure 4 -Simplified Structure Polygons Figure 5 - USGS Topographic Map

42

Section 4 Appendix - Digital Processes Technical Specifications

43

4.0.0 Technical Appendix This section of the report details the process steps, functions, specific tools and their variables that

were used in this research. The demonstration maps follow these particular specifications, though

certain figures found in this report may reflect different variable values or processes. Input data and

output data names have been simplified for sake of clarity. .

4.1.0 Generation Process: Urban Tint from Road Data

1.) Create Input: This step extracts the usable features from the Census TIGER data. Process: A series of functions and tools were used to perform this task. Input: US Census TIGER Data Tool/Process: Select if MTFCC = x (x = MTFCC) S1100 (Primary Roads) S1200 (Secondary Roads) S1400 (Local Roads) S1730 (Alley) S1740 (Private Road) Tool/Process: Export Selected Output: Transportation Input 2.) Classification Weight: This step creates a new field in the road data and assigns feature weights using the field calculator. Process: A series of functions and tools were used to perform this task. Input: Transportation Input Tool/Process: Add new field Field Name: weight Field Type: double Tool/Process: Select if MTFCC = x Tool/Process: Field Calculator (If x Then = v) x = MTFCC , v = code weight value Calculate Value = v MTFCC code weights: Primary Roads - S1100 = 0 Miscellaneous- S1730,1740 = 1 Secondary Roads - S1200 = 2 Local Roads – S1400 = 3 Output: Transportation Weighted 3.) Create Density Raster: This step creates a density raster from the road data using feature weights. Tool: Line Density Input: Transportation Weighted Weighted on MTFCC 1-3 Population: weight value Output Cell Size:10 meter Search Radius: 120 meter Area units: Acres Output: Road Density

44

4.) Focal Statistic: This step processes the density raster through a focal statistics process. Tool: Focal Statistics Input: Road Density Statistics Type: Mean Neighborhood Shape: Annulus Neighborhood Size: Inner Radius 2 cell, Outer Radius 15 cell Output: Road Focal 5.) Reclassify: This step reclassifies the focal raster into two integer classes. Tool: Reclassify Input: Road Focal Method: Manual Class Values: 2 classes, excluding '0' values Value 1 < 1.9 Value 2 > 1.9 Output: Reclassified Road Raster

4.1.1 Generation Process: Urban Tint from Structure Data

1.) Calculate Polygon Area: This step creates a field in the structure polygon data, values are populated by the field calculator. Process: A series of functions and tools were used to perform this task. Input: Structure Polygons Tool/Process: Add new field Field Name: Area Field Type: float Tool/Process: Field Calculator Field: Area Calculate Area = Square Feet Output: Structure Polygons 2.) Intersect: This step assigns the calculated area to points that have a corresponding structure polygon. Process: A series of functions and tools were used to perform this task. Tool/Process: Intersect Input 1: Structure Point Input 2: Structure Polygon Output: Structure Point Weight 2.) Create Weight Value: This step assigns the average area (7500) and the point weight value to all points. The calculated area (square feet) divided by 7500 and values > 1 multiplied by 2.5. Input: Structure Point Weight Tool/Process: Select Field Name: area Select if 'Null' Tool/Process: Field Calculator Field: Area Calculate Selected Features

45

Values: = 7500 Tool/Process: Field Calculator Field: Weight Value : area/7500 Tool/Process: Select Field Name: weight Select if > 1 Tool/Process: Field Calculator Field: weight Calculate Selected Features Value: weight x 2.5 3.) Create Density Raster: Tool: Point Density Input: Structure Point Weight Population: weight Output cell size:30 Search Radius: 30 Area units: Acres Output: Structure Density 4.) Reclassify: Tool: Reclassify Input: Structure Density Method: Manual Output: Reclassified Values: 2 class, excluding '0' values Value 1 < 2.0 Value 2 > 2.0 Output: Reclassify Structure Raster

4.1.2 Generation Process: Urban Tint from Road and Structure Data

1.) Create Road Input: This step creates the classified road data raster. Complete Section 4.1.0 Steps 1-3 Tool: Reclassify Input: Road Focal Method: Natural Breaks (Jenks) Output: Reclassified Values: As defined, 4 class, excluding '0' values. Output: Reclassify Road 2 2.) Create Structure Input: This step creates the classified structure data raster. Complete Section 4.1.1 Steps Tool: Reclassify Input: Structure Density Method: Natural Breaks Output: Reclassified Values: As defined, 6 class, excluding '0' values.

46

Output: Reclassify Structure2 3.) Add Rasters: This step adds the classified structure and road data rasters and produces a combined raster. Tool: Plus Input 1: Reclassify Structure2 Input 2: Reclassify Road 2 Output: Combined 4.) Reclassify: Tool: Reclassify Input: Combined Method: Manual Output: Reclassified Values: 2 class, excluding '0' values Value 1 < 4 Value 2 > 5 Output: Reclassify Combined 5.) Cartographic Process: Complete 4.1.3 4.1.3 Cartographic Process 1.) Raster to Polygon: This step converts the reclassified raster to a polygon dataset. Tool: Raster to Polygon Input: Reclassify Output: Road/Structure Polygon 2.) Delete Polygons Process: A series of functions and tools were used to perform this task. Input: Road/Structure Polygon Tool/Process: Select Features Field Name: Grid Select if 1 Tool/Process: Delete Selected Tool/Process: Add new field Field Name: Area Field Type: float Tool/Process: Field Calculator Value = Calculate Acres Tool/Process: Select Features Field Name: Area Select if < 40 Tool/Process: Delete Selected Output: Road/Structure Less40Acres 3.) Aggregate Polygon Input: Less40Acres Output: Aggregate Aggregation Distance: 660 feet Min Area: 40 acres Min Hole Size: 3000 acres

47

4.) Smooth Polygon Tool: Smooth Polygon Input: Aggregate Smooth Algorithm: Paek Smoothing Tolerance: 120 meter Output: Urban Tint 4.2.0 Simple Structure Symbology Process 1.) Add Rotation Angle: This adds the necessary rotation angle.

Tool: Near Input: Structure Point Data Near Feature: Roads Search Radius:1000 ft Option: Include Angle Output: Structure Point Data

2.) Rotate Angle: This is done in ArcMap. Symbology/Advanced/Rotation

Field: Near_Angle Rotation Style: Arithmetic

4.2.1 Simplification Process

1.) Simplify Structure: This tool simplifies the structure polygon.

Tool: Simplify Building Input: Structure Polygon Data Tolerance: 80 feet Simplification Tolerance: (none) Output: Structure Polygon Data

usgs next generation topographic map - graphics research

Documents